How Corruption Impacts Poverty in Developing Countries? The Role of Education

(1)

HOW CORRUPTION IMPACTS POVERTY IN DEVELOPING

COUN-TRIES? THE ROLE OF EDUCATION

Carlota Campos Castro

Dissertation

Master in Economics

Supervised by

Maria Isabel Gonçalves da Mota Campos

(2)

ii

Acknowledgments

To Professor Isabel Mota I am very grateful for all the patience, dedication and all she has taught me throughout this year. The exigency and professionalism with which she accompanied me, along with all the constructive advices that made me go further, is some-thing that I will always take into consideration.

To my family and friends for all the love, support, understanding and confidence they have given me.

To the volunteer project of which I am part, o Grão, for teaching me that the true meaning of being present is to be present, for making me look to the small things, for teach-ing me to give myself totally to everythteach-ing I do without expectations and for teachteach-ing me to put as much as I am in the least that I do.

I will be forever grateful.

Para ser grande, sê inteiro: nada Teu exagera ou exclui.

Sê todo em cada coisa. Põe quanto és No mínimo que fazes.

Assim em cada lago a lua toda Brilha, porque alta vive

Ricardo Reis, in "Odes" Heterónimo de Fernando Pessoa

(3)

iii

Abstract

The main objective of this dissertation is to analyse the impact of corruption on poverty in developing countries and how does education influence this relationship. Particu-larly, we aim to answer the following research questions: How does corruption influence poverty in developing countries and particularly, in low-income countries? How does educa-tion affect poverty? What is the role of educaeduca-tion in the control of corrupeduca-tion and how is it related to poverty? Literature on the relationship between corruption and economic growth is considerable but studies that relate poverty and corruption are quite few and the issue of education is almost inexistent.

We start with a literature review on the relationships between poverty and corruption, as well as with the relationship between corruption and education. Thereafter, an economet-ric model is estimated using panel data analysis and considering a sample of 81 low and middle-income countries for the period 1998-2017. Our model uses the Poverty Headcount Ratio as dependent variable and the Corruption Perception Index and education enrolment and expenditures in education as independent variables, after controlling for other variables. We also consider a subsample with the low-income countries, and subsamples according to the education level.

It is possible to conclude that corruption has a strong negative impact on poverty. However, the variable loses significance when considering the low-income countries. Our second conclusion is that education also impacts poverty negatively, having an important role in poverty reduction. Thirdly, we find that the impact of the control of corruption on pov-erty is higher for developing countries with more investment on education and for countries with lower school enrolment in primary.

As a result, countries should invest in education to make people more aware of the consequences of corruption, how to detect it and how to avoid it in order to promote pov-erty alleviation in developing countries.

JEL-codes: D73; I32; I25; O15; C23

(4)

iv

Resumo

O principal objetivo desta dissertação é analisar o impacto da corrupção na pobreza nos países em desenvolvimento e como a educação influencia essa relação. Particularmente, pretendemos responder às questões: Como é que a corrupção influencia a pobreza nos países em desenvolvimento e, particularmente, nos países de baixo rendimento? Como é que a edu-cação afeta a pobreza? Qual é o papel da eduedu-cação no controlo da corrupção e como ela está relacionada com a pobreza? A literatura considera a relação entre corrupção e crescimento económico, mas apenas alguns trabalhos que relacionam pobreza e corrupção e a questão da educação é quase inexistente.

Começamos com uma revisão de literatura com as contribuições da relação entre pobreza e corrupção, e da relação entre corrupção e educação. Posteriormente, um modelo econométrico é estimado utilizando dados em painel para uma amostra de 81 países no pe-ríodo 1998-2017, onde consideramos o Índice de Incidência de Pobreza como variável de-pendente e o Índice de Perceção da Corrupção e o registro escolar de educação e gastos em educação como variáveis independentes. Consideramos também subamostras com os países de baixo rendimento e subamostras de acordo com o nível de escolaridade.

Concluímos que a corrupção tem um impacto negativo sobre a pobreza, perdendo esta variável significância nos países de baixo rendimento. A nossa segunda conclusão é que a educação afeta negativamente a pobreza, tendo um papel importante na sua redução. Em terceiro lugar, descobrimos que o impacto do controlo da corrupção sobre a pobreza é maior para os países em desenvolvimento com mais investimentos em educação e para os países com menos matrículas em escolas primárias.

Como resultado, os países deveriam investir na educação para tornar as pessoas mais conscientes das consequências da corrupção, como detetá-la e evitá-la, para promover a re-dução da pobreza nos países em desenvolvimento.

Códigos-JEL: D73; I32; I25; O15; C23

(5)

v

List of tables

Table 1 – Types of corruption ... 10

Table 2 - Mechanisms through which corruption affects poverty ... 14

Table 3 – Structure of the sample by income level ... 22

Table 4 – Summary Statistics ... 27

Table 5 – Correlation Matrix ... 29

Table 6 – Specification and diagnosis tests: Full sample... 31

Table 7 – Corruption and Education on Poverty, 1998-2017: Full sample... 33

Table 8 – Specification and diagnosis tests: Low income subsample ... 36

Table 9 – Corruption and Education on Poverty, 1998-2017: Low-income subsample ... 38

Table 10 – Corruption and Education on Poverty, 1998-2017: ... 39

Group I (EDUPRI > 101.7) and Group II (EDUPRI < 101.7) ... 39

Table 11 – Corruption and Education on Poverty, 1998-2017: ... 40

Group III (EDUSEC > 62.5) and Group IV (EDUSEC < 62.5) ... 40

Table 12 – Corruption and Education on Poverty, 1998-2017: Group V (EXPEDUC > 4.21) and Group VI (EXPEDUC < 4.21) ... 41

Table 13 – Corruption and Education on Poverty, 1998-2017: Interaction ... 43

(7)

vii

List of figures

Figure 1 – Poverty HCR in low- and middle-income countries, 1998-2017 ... 23 Figure 2 – CPI in low- and middle-income countries, 1998-2017 ... 24

(8)

1

Chapter 1. Introduction

Even after the Second World War, poverty is still a problem for developing and un-derdeveloped countries (Yildiz, 2017). Extreme poverty, weak service delivery, high levels of corruption and the low standards of living of people are some of the characteristics in the world of poor (Madinda, 2014).

According to Nwankwo (2014), corruption has existed since pre-biblical times, both in developed and developing nations. As reported by the Transparency International1_,

cor-ruption is one of the major challenges in the world and it affects the functioning of govern-ments and public policies, leading to a bad allocation of resources, damaging the private sector and harming the poorest. Corruption is defined by Deininger and Mpuga (2005) as "the abuse of public power for private benefit, as a key constraint to efficient allocation of economically valuable resources, effective provision of public goods and services, and peo-ple's confidence in the state and the legal system" (Ibid, p.5). The authors point out that corruption is highly exposed when there is high level of economic integration between coun-tries and easy access to information – reducing the scope of discretionary action made from the public officials and improving the economic efficiency, playing a main role in the devel-opment process. The need for transparency in the allocation of public expenditures is crucial if the aim of investments in public goods is to increase the economic growth of developing countries.

Corruption affects poverty, and, in turn, poverty can impact corruption. Chetwynd, Chetwynd and Spector (2003) and Yildiz (2017) point out that corruption has consequences on governance and economic factors, and this increase poverty. Additionally, authors such as Negin, Rashib and Nikopour (2010) argue that corruption can cause poverty. As cited by Annan (2004) “corruption hurts the poor disproportionately diverting funds intended for development, undermining a governments ability to provide basic services, and feeding ine-quality and injustice and discouraging foreign aid and investment” (Ibid, p.iii). In countries affected by corruption, there is weak trust on public institutions, the quality of public ser-vices is not very good – the expenditures on health and education are not a priority – and this increases the levels of poverty (Chetwynd et al., 2003). Literature also states that poverty can impact corruption. In Mauro (1998)’s article, there is evidence that poor countries have

(9)

2 more corrupt activities because of the difficulty of allocating resources that these countries face. Also, Unver and Koyuncu (2016) argue that countries with higher levels of poverty face higher levels of corruption.

Chetwynd et al. (2003) state that poverty is multidimensional and it can be related with low levels of income, weak levels of education and health, vulnerability and, also pow-erlessness. Siddique, Shehzadi, Shaheen, & Manzoor (2016) point out that poverty and edu-cation are influenced by economic growth, governments and institutions, since governments and institutions are the main influences in the poverty reduction and the in quality of edu-cation. Eicher, García-Pẽ alosa and Ypersele (2009) sustain that corruption can negatively affect education, but education also influences corruption. In fact, corruption leads to the decrease of available income and investment in education and education generates high out-put, corruption rents and increases the risk of corrupt politicians being detected and pun-ished.

This dissertation aims to examine the impact of corruption on poverty in developing countries and how education mediates this relationship. Particularly, this study intends to answer the following questions: How does corruption influence poverty in developing coun-tries and particularly, in low-income councoun-tries? How does education affect poverty? What is the role of education in the control of corruption and how is it related to poverty?

To achieve our goals, we estimate through OLS an econometric model by using panel data analysis to understand the influence of corruption and education on poverty, after con-trolling for other variables. Our model uses the Poverty Headcount Ratio as dependent var-iable and the Corruption Perception Index and education enrolment and expenditures in education as independent variables, after controlling for other variables. We consider a da-taset with 81 countries for the period from 1998 to 2017. The data used is retrieved from the Transparency International and the World Bank (2019). Using the classifications of the World Bank (2019), we consider the low-income and middle-income countries and also a subsample with only the low income countries. In order to analyse the role of education in the relationship between corruption and poverty, we also divide our sample in subsamples according to the education level.

This topic deserves attention for several reasons. For academic purposes, since to the best of our knowledge, in spite of existing a significant research on the topic of corruption and economic growth, there are very few works relating corruption and poverty, and the

(10)

3 issue of education is almost absent. Azward (2018) studies the relationship between corrup-tion and poverty focusing on Indonesia. Although not including educacorrup-tion as an independent variable, the author compares the social spending on education with the Corruption Percep-tion Index (CPI). We go further and include educaPercep-tion, measured by enrolment rates and by government expenditures on education, as explaining poverty, and to study if the effect of corruption on poverty is affected by education. In addition, for social and political reasons, as eradicating poverty is one of the greatest challenges facing humanity.2

This dissertation is organized as follows. After this Introduction, chapter 2 starts with the definition of the key concepts – poverty and corruption. Chapter 3 explains the econo-metric model estimated, as well as the data. Chapter 4 shows and discuss the results. Chapter 5 concludes and offers new paths for future research.

2 _United _Nations _(2018), _Sustainable _Development _Goals, https://www.sustainabledevelop-ment.un.org/?menu=1300 (accessed on 12/11/2018)

(11)

4

Chapter 2. Corruption and Poverty: a literature review

In this chapter we introduce some concepts and measures of poverty and corruption. We will start with the definition of poverty from the monetary perspective, either absolute or relative, and then proceed with Sen’s capability approach. On the second part of the chap-ter we will discuss the concept of corruption and its measurement, either considering per-ception or non-perceptual measures.

2.1 Poverty: concepts and measures

For many years, there has been an effort to define poverty and to find ways of accu-rately measure it. In this section, we will present some distinct perspectives of poverty, as well as a list of procedures the authors use to measure it.

2.1.1 Concepts

Poverty can be defined as the lack of resources needed to meet the basic needs of an individual or family (Fields, 1994). In Mabughi and Selim (2006) paper, the authors aim to provide a definition of poverty as social deprivation. As pointed by them, absolute pov-erty can be defined by “the subsistence below a minimum, socially acceptable living condi-tion, established based on nutritional requirements and other essential goods” (Ibid, p. 184).

Mabughi and Selim note that Rowntree's in 1901 was one of the major contributors to the definition of absolute poverty. Absolute poverty is accounted by the number of indi-viduals living below a poverty line, an income or consumption threshold below which pop-ulation is considered poor (Mabughi & Selim, 2006). Additionally, Yildiz (2017) considers that absolute poverty is related to human needs and their deprivation. The author claims that this deprivation is associated with the lack of food, sheltering, safe drinking water, health services, access to information, sewerage facilities and education.

In the definition of the poverty line, families and individuals are considered poor when they are below a consumption threshold, usually defined as a minimum level of social well-being in terms of calories per day - 2200 calories (Mabughi & Selim, 2006). Moreover, poverty lines can also be defined through income. Since 2016, the World Bank defines ex-treme poverty as individuals living with less than US $ 1.9 per day (PPP).3

3 _{http://www.worldbank.org/en/news/feature/2016/06/08/ending-extreme-poverty} _(accessed _on 28/10/2018)

(12)

5 To assess the extent of poverty, we may count the number of people below the pov-erty line or the extent of the resource gap (Fields, 1994). Fields refers to a question that leads to the debate about the existence of a single poverty line for all developing countries or whether the poverty line is defined for each country. For him, there are advantages in using a common and internationally comparable poverty line. However, there are countries with well-established national poverty lines that do not accept the common poverty line estab-lished by the World Bank (WB) claiming that they are poorer than what the WB refers to (Mabughi & Selim, 2006). Since 2015 the World Bank reports two higher-value poverty lines: $3.20 (low-middle income) and $5.50 (upper-middle income) per day.

A different approach to poverty is the relative one. Mabughi and Selim (2006) define relative poverty as “measured in terms of judgments by members of a particular society by what is considered a reasonable and acceptable standard of living” (Ibid, p. 186). The authors point out that Townsend in 1979 makes a major contribution to relative poverty, when he argues that individuals with lack of access to resources that are necessary for normal living conditions may be in poverty. Therefore, the authors consider that relative poverty is de-pendent on social expectations and living standards and consequently some luxuries may be assumed as necessities. The concept of relative poverty is represented in two dimensions: poverty with low incomes and poverty as a private lifestyle (Townsend, 1979) (cfr. Mabughi & Selim, 2006). According to the authors, Townsend considers that income cannot be the only variant in the definition of poverty, including in its conclusions in 1979, variables such as work, friends or housing conditions that are important to keep a stable standard of living.

Misturelli and Heffernan (2008) state that literature offers numerous definitions of poverty. Besides the income-based concept above mentioned, poverty may also be defined though the capability or multidimensional perspective.

According to Sen and Anand (1997), poverty is the deprivation of the basic necessi-ties and the lack of opportuninecessi-ties for human lives. As so, the use of a multidimensional approach is required for an appropriate indicator of human poverty. According to Atkinson (2003) the deprivation is multidimensional. Based on several studies, Misturelli and Heffernan (2008) affirm that the multidimensional concept of poverty has grown from the social construction of poverty, and it can be seen in social, cultural and political standards at the community level. The Multidimensional Poverty Index (United Nations Development

(13)

6 Programme, 2018)4 measures poverty and compares it internationally, focusing on real

dep-rivation. This index presents three dimensions: health, education and living standards. Regarding the capability approach defined by Sen (1985), poverty is not measured by what one has but instead by what can be done with it. Misturelli and Heffernan (2008) defend that the concept of poverty that refers only to the lack of material things is limited and partial and one should also analyse the ability to achieve basic skills.

According to Sen (1999) freedom is essential for the general well-being and the re-duction of deprivation is fundamental for development. Actually, poverty can be decreased by political freedom and stability in opportunities, empowerment, capabilities and security (Negin et al., 2010). Nafziger (2006) describes the theory of Sen and state that “unfreedoms include hunger, famine, ignorance, an unsustainable economic life, unemployment, barriers to economic fulfilment by women or minority communities, premature death, violation of political freedom and basic liberty, threats to the environment, and little access to health, sanitation, or clean water.” (Ibid, p.45). Indeed, the author states that some components of development such as freedom of exchange, labourcontract and social opportunities are also meaning to achieve development and freedom.

Sen's welfare theory is based on the capabilities of individuals and not on their indi-vidual achievements (Nafziger, 2006). The author also states that Sen uses a reduced number of basic functionings necessary for well-being. According to Nafziger, some of the basic functionings presented as examples by Sen are “being adequately nourished, avoiding prem-ature mortality, appearing in public without shame, being happy, and being free” (Ibid, p.178).

2.1.2. Measures

There are different measures of poverty, such as, the Headcount Ratio (HRC), the Poverty Gap Ratio (PGR) and the Squared Poverty Gap, with different characteristics and the limitations.

Mabughi and Selim (2006) use the Headcount Ratio (HCR) to measure the propor-tion of populapropor-tion below the monetary poverty line. This index has some problems referred by the authors. The HCR only counts the number of poor people and not how much of the individuals incomes are below the poverty line. Also, it does not reflect the depth of poverty nor how poor individuals are harmed by the relative severity of poverty, as also stated by Chong, Gradstein and Calderon (2009).

(14)

7 The Poverty Gap Ratio (or income-gap approach) overcomes some these limitations as it measures the amount of income necessary to move the poor out of the poverty line (Nafziger, 2006). Still, Mabughi and Selim argued that the Poverty Gap Ratio does not give enough importance to the distribution of income among the poor, since it does not account for income inequality among the poor. Chong et al. (2009) agree with this limitation, adding that the index remains unchanged when the poor make monetary transfers to the less poor.

As so, the Squared Poverty Gap gives more relevance to the poverty gaps, becoming an indicator of the severity of poverty. The Squared Poverty Gap reflects the monetary transfers between the poor below the poverty line and the poorest. For Alvi and Senbeta (2011), this index brings more weight to the poorest of the poor.

Sen's article (1976) presents two distinct issues: determining how many are the poor relatively to the total population and the construction of a poverty index given the existing information about the poor. Regarding the first one, the problem refers to the poverty criteria chosen, namely the choice of the "poverty line", and the determination of individuals that are below the "poverty line" and those who are not. The procedure used by the author – the Headcount Ratio – to combat the second problem, aims to account for the number of poor and their respective percentage of the total population of poor. However, Sen (1976) also presents the limitations of the Headcount Ratio and the Poverty Gap Ratio. The first one has problems measuring the fall in poverty per person and the second one has difficulties in measuring the numbers involved. Neither of them presents enough information on the dis-tribution of income of the poor.

2.2 Corruption

For the past few years, different authors try to define corruption and to find a way to truthfully measure it. However, corruption is hard to define and even harder to measure because corruption can be found in different ways and is that one thing that everyone tries to hide. We will start by describing some of the perspectives from different authors when they try to define a concept for corruption and then we will present the procedures used to measure it.

2.2.1. Concepts

Yusuf (2012) states that corruption can be defined as the abuse of power that an individual withdraws from his position with the goal of achieving individual gains. Jain (2001)

(15)

8 states that this power is used over the allocation of resources from different types of agents – the political elite, the administrators and the legislators. Tanzi (1998) considers that in some cases of corruption “the abuse of public power is not necessary for one’s private benefit but it can be for the benefit of one’s party, class, tribe, friends, family, and so on” (Ibid, p.8). The Transparency International5 states that “corruption can be classified as grand, petty and

po-litical, depending on the amounts of money lost and the sector where it occurs”.Additionally, the Transparency International Australia6 describes corruption as “one of the greatest

chal-lenges of the contemporary world. It undermines good government, fundamentally distorts public policy, leads to the misallocation of resources, harms the private sector and private sector development and particularly hurts the poor”. Undermining good government – gov-ernment with low levels of inequality, diverse people and that are good for economic devel-opment, with security of property rights, high quality of bureaucracy, and effective spending, among other things – can negatively impact the country since it is important for its growth and its economic development (La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1999).

The paper of Lalountas, Manolas and Vavouras (2011) focus on the public sector corruption.They characterize corruption as access to public goods and services, bribes, ille-gal hiring in the public sector, among other things. Zhao, Kim and Du (2003) (cfr. Unver & Koyuncu, 2016, p.4), state that “corruption acts as a major deterrent to perfect competition and creates political instability and social issues”.

La Porta et al. (1999) states that corruption is related with bureaucratic discretion, since the delays caused by it generate opportunities to take bribes. Also, in countries with high levels of corruption, politicians earn huge wages and collect bribes due to their power. For Blackburn and Powell (2011), corruption can be defined as misappropriation of public funds and in their results they claim that the larger the corruption, the higher the rates of monetary growth. Tanzi (1998) states that corruption does not only result in bribes, since it can be found in individuals that abuse of the power gained from their public position, such as, individuals that go on vacations but claim being sick.

Different authors like Amundsen (1999), Jain (2001), Perera and Lee (2013) Shleifer

5 _{https://www.transparency.org/what-is-corruption#define}_{(accessed on 7/11/2018)} 6 _{http://transparency.org.au/mission-statement/}_{(accessed on data 24/10/2018)}

(16)

9 and Vishny (1993), Tanzi (1998), Truex (2011), Yildiz (2017), Yusuf (2012) and the Trans-parency International7_{identified different types of corruption. In Table 1 we offer a}

cate-gorization of corruption, aiming to synthetize different perspectives and classifications. First, we identify political corruption, which is the manipulation of policies, organ-izations and rules in the allocation of resources, it involves political decision-makers and it occurs in the highest positions of the political system. We may also include in this category legislative corruption, defined as the bribes that interest groups pay to legislators in order to influence the voting results; lobbying, where an activity can influence the decisions of politics and bureaucrats; and vote trading (form of political corruption), where electors buy votes from voters.

Another category of corruption is grand corruption, defined as economic policies that politicians do by abusing their power, in order to achieve their personal interests and not only the interests of the population.

Also, petty corruption is one of the categories of corruption. It consists in the abuse of power made by public officials in their low and mid-level positions that ask people for something in the exchange of access to public services or goods such as schools, hospi-tals, police and other agencies. In this category we include bureaucratic corruption: the bureaucratic are corrupt and their actions are reflected in the relationship with the political elite or the population, making individuals pay bribes to have access to public services or bureaucratic procedures.

We then consider economic corruption as a category of corruption, where it is included government corruption, represented by government officials. Other types of cor-ruption are involved in this category, such as, rent-seeking, for example a tariff, license, or quota; nepotism, when public officials favour people they know either friends, family or other politicians in an illegal way; gift vs cash, that refers to when a tax collector receives some gift/cash from the shopkeeper with the objective of not having to pay taxes; public vs private, where a job is offered to a friend or family even if other candidates were more qualified or where a government employee or a businessman receive a gift in order to receive a construction contract either public or private; favouritism, referring to situations where the people involved are friends and because of that they receive something easier; develop-mental corruption, found in East Asia, and it happens when public officials receive a slide of profits from political activities in exchange for protection of the private industry and the

(17)

10 provision of resources to it; and degenerative corruption, found in Africa, the Caribbean and Latin America, that corresponds to looting of treasure or extortions to private properties made by public officials.

Finally, the last category is financial corruption that incorporates bribery, defined as a payment made to an individual or institution in exchange of illegal actions; insider trad-ing, that happens when investors have access to insider information prior to public an-nouncements and use this information to make investments in the capital markets, giving advantages to other investors; and money laundering, which can be defined as a commis-sion of an act in order to undercover the corruption.

Table 1 – Types of corruption

Types of corruption Political

Political corruption (Amundsen, 1999; Transparency International8_),

Lobbying (Yildiz, 2017), Vote trading (Yildiz, 2017), Legislative cor-ruption (Jain, 2001)

Grand Grand corruption ( Jain, 2001; Transparency International9_),

Petty Petty corruption (Transparency International

10_{), Bureaucratic}

cor-ruption (Jain, 2001)

Economic

Government corruption (Shleifer and Vishny, 1993), Rent-seeking (Yildiz, 2017), Nepotism (Yildiz, 2017), Gift vs cash (Truex, 2011), Public vs private (Truex, 2011), Favouritism (Truex, 2011), Develop-mental corruption (Perera, 2013), Degenerative corruption (Perera,

2013)

Financial Bribery (Yildiz, 2017), Insider trading (Yildiz, 2017), Money launder-ing (Yusuf, 2012)

Source: own elaboration

As stated in the literature (e.g. Tanzi (1998); Jain (2001)) as the definition of corrup-tion also measuring corrupcorrup-tion is really hard. However, it is possible to find some measures, such as, Business International Corporation (BI), Bribe Payers Index, Corruption Percep-tions Index (CPI) and World Bank Group’s Worldwide Governance Indicators (WGI) ( Mauro, 1995; Jain, 2001; Heywood & Rose, 2014).

8 _{https://www.transparency.org/what-is-corruption#define}_{(accessed on 26/10/2018)} 9 _{https://www.transparency.org/what-is-corruption}_{(accessed on: 28/12/2018)} 10 _{https://www.transparency.org/what-is-corruption}_{(accessed on: 28/12/2018)}

(18)

11 2.2.2. Measures

According to Jain (2001) and Tanzi (1998), it is very difficult to measure corruption, because it is very hard to measure something that people try to hide. However, in his research, there are some measures identified such as Business International Corporation (BI) and Bribe Payers Index. Regarding BI, published by the Economist Intelligence Unit, it refers to a pri-vate firm that computes indices of the corrupt countries based on the risk and efficiency factors for the period of 1980-1983 and sell it to banks, companies and other private inves-tors (Mauro, 1995). Mauro (1995) uses BI and defines corruption as “the degree to which business transactions involve corruption or questionable payments” (Ibid, p.684). On the other hand, the Bribe Payers Index is produced by the Transparency International, an organ-ization that focuses on identifying bribes in the entire world. Also, Heywood and Rose (2014) believe it is difficult to measure corruption because there is no authorized agreed definition of what corruption is. However, they point some measures such as the Corruption Percep-tions Index (CPI) from Transparency International, the World Bank Group’s Worldwide Governance Indicators (WGI) and, like Jain (2001), the Bribe Payers Index. According to Tanzi (1998), measuring corruption is impossible, the only possible thing is to measure the perceptions of corruption. Heywood and Rose (2014) state that the CPI and the WGI are perceptions measures to combat corruption, since they focus in public sector bribes and misappropriation of public funds for private purposes. According to the authors, these measures of perception are correlated with measures of corruption based on experience – measures that relied on honest reports from individuals or reality –, although this relationship is not linear. So, the authors argue that perceptual measures correspond to absolute levels of corruption within countries, which means that even if corruption is the same per person, larger countries will have higher levels of corruption. There are some limitations mentioned by the authors about these measures since perceptions may not demonstrate reality. In fact, these measures are not able to differentiate the corruption types or their distinction between sectors, thus presenting a disadvantage.

Heywood and Rose (2014) suggest non-perceptual measures of corruption. These measures are based on existing infrastructures compared to monetary investments in regions, to perceive the absence of physical infrastructures in each region. Also, the non-perceptual measures may be founded on levels of conviction of public sector officials for corruption offenses. Some mathematical models were also implemented as a measure of the study of

(19)

12 corruption, especially in cases of electoral fraud. According to the authors, both non-percep-tive measures and percepnon-percep-tive measures face problems with the concept of corruption, the identification of different types of corruption, and with dealing with specific cases of corrup-tion or possible solucorrup-tions to their resolucorrup-tion.

2.3 Poverty and corruption: main insights from the literature

Corruption influences economic development, captured by economic growth, or by other variables such as poverty and income inequality. Before analysing the relationship be-tween poverty and corruption, it is relevant to understand the influence of corruption on economic growth.

Yildiz (2017) states that corruption has impacts on economic growth because it de-creases the quantity and quality of public services, such as the expenditures on infrastructures, tax revenues and human capital accumulation. Additionally, corruption influences the devel-opment of a country, and as Jain states it “seems to affect the level of investment, entrepre-neurial incentives, and the design or implementation of rules or regulations regarding access to resources and assets within a country” (Ibid, 2001, p.72).

According to Mauro (1995), corruption reduces investment and economic growth because of rent-seeking – there are bad allocation of resources that can be a barrier to new private investments, therefore decreasing economic growth. He states that there are many economists who argue that the obstacles to investment, entrepreneurship and innovation are due to the poor functioning of governments. The author finds a negative relationship be-tween corruption and investment, and in turn, economic growth. However, if countries make the integrity and efficiency of their bureaucracy better, corruption may induce increases in the countries investment rate and in turn in GDP. Mauro says that rich countries have better government institutions than poor countries. However, he also claims that corrupt countries, such as Thailand, have a very positive economic growth, showing the ambiguity of this ar-gument. Moreover, according to Leff (1964), corruption can benefit economic growth or it can worsen it depending on the context and perspective analysed. Corruption may help eco-nomic growth when the rate of investment is increased. Also Bardhan (1997), who follows the study of Leff in 1964, states that in developing countries, corruption benefits economic growth and increases efficiency. The author uses the example of Philippines, where price discrimination is practice by corrupt bureaucrats with a different time preference. Another example is when bureaucrats take advantage of corruption in the allocation of contracts or

(20)

13 licenses benefiting the interested parties and thus increasing the efficiency.

2.3.1. Corruption and Poverty

According to Mabughi and Selim (2006), poverty is a social deprivation, and is influ-enced by several variables. The authors identify the location (people that lives in areas far from markets, health centres and schools for children), education (people that has less than the basic education) and health (lack of access to health services) as main determinants of poverty. In addition, Gupta, Davoodi, and Alonso-Terme (1998) considers that income ine-quality impacts the poor. Chetwynd et al. (2003) point out that poverty is influenced by the levels of education and health. Rahayu and Widodo (2012) also include powerlessness and vulnerability. They consider that these four variables decrease poverty and invite corruption. Negin et al., (2010) point out that rural areas strongly depend on natural resources that leads to risk environment, the high transactions costs and lack of access to social and physical infrastructures that are impacted by the low density of the population and their geographical constrains, being associated with higher poverty rates.

Many authors consider that poverty can be also affected by corruption. In Appendix 1, we list some empirical studies that focus on the relationship between corruption and pov-erty.

According to Negin et al. (2010) “corruption is a cause of poverty and a barrier to successful poverty eradication” (Ibid, p.2). Chetwynd et al. (2003) and Negin et al. (2010) state that corruption in the public sector worsens poverty – causing low income, weak health and education status and the countries became more vulnerable to shocks –, decelerating the economic growth in these countries.

According to Yildiz (2017), who agrees with Chetwynd et al.¸ the relation between corruption and poverty is an indirect one because the consequences of corruption on pov-erty are on economic and governance factors first and then this will increase povpov-erty.

In Table 2, we summarize mechanisms through which corruption affects poverty. As we can see, corruption affects the economic and political stability, it affects the income dis-tribution and the government effectiveness, it worsens the public sector and reduces the investments in education and health, and economic growth suffers impacts.

(21)

14

Table 2 - Mechanisms through which corruption affects poverty

Mechanisms References

Corruption ➔ Income distribution

Decreases the income growth rate of bottom 20% of the population

Gupta et al. (2002), Chetwynd et al. (2003), Negin et al. (2010), Tebaldi and Mohan

(2010)

Reinforces income inequality Rahayu and Widodo (2012) Corruption

➔ Institu-tions

Poor political stability Tebaldi and Mohan (2010) Reduces government effectiveness Tebaldi and Mohan (2010) Worsens democratic accountability Perera and Lee (2013)

Weakens institutional quality Perera and Lee (2013) Corruption

➔ Eco-nomic

con-juncture

Decreases economic stability Gupta et al. (2002) Lowers economic growth

Chetwynd et al. (2003), Negin et al. (2010), Rahayu and Widodo (2012), Azward

(2018)

Corruption ➔ Public

services

Increases prices of public goods Rahayu and Widodo (2012) Worsen the quantity and quality of

pub-lic services Rahayu and Widodo (2012)

Makes it difficult to have access to pub-lic services

Rahayu and Widodo (2012), Justesen and Bjørnskov (2014)

Distorting public expenditure allocation Rahayu and Widodo (2012) Lower social spending of governments

on education and health

Mauro (1997), Tanzi (1998), Chetwynd et al. (2003), De la Croix and Delavallade (2009), Eicher et al. (2009), Azward (2018) Source: own elaboration

Gupta, Davoodi, and Alonso-Terme (2002) analyse the effects of corruption on in-come inequality and poverty. They use inin-come growth of the bottom 20% as the dependent variable, and independent variables such as: natural resource abundance, initial income of the poor, initial secondary schooling, education inequality, initial distribution of assets, social spending, and corruption. The authors find that corruption impacts poverty, since an in-crease on the growth rate of corruption dein-creases the income growth of the poor. They conclude that resources allocation, economic stability and income distribution are character-istics impacted by corruption, therefore, poverty is also affected.

(22)

15 and poverty. The authors state that poverty does cause corruption, but corruption directly affects the poor because it increases the prices of public services, leads to smaller quantity and quality of public services, and makes it difficult to have access to public services – health care, education, water, sanitation, and others. They conclude that corruption has indirectly impacts on poverty, by decreasing economic growth, reinforcing income inequality and dis-torting public expenditure allocation.

Dincer and Gunalp (2008) investigate corruption, income inequality and poverty in the US. They estimate a model where poverty, measured as a percentage of people with income below the poverty threshold, is the dependent variable and corruption was measured as the number of government officials criminals convicted for corruption. Regarding the independent variables, they use poverty, income, government policies, education, union, and unemployment. The authors find that corruption leads to an increase in poverty levels, where variables such as education, income, union, and unemployment are significant. According to them, poverty can be affected by corruption and directly and indirectly through income ine-quality.

There are also studies that adopt a broader approach and focus on the influence of institutional variables on poverty, including corruption. Tebaldi and Mohan (2010) analyse institutions and poverty and their relationship. They consider poverty as the dependent var-iable measured by the percentage of population living on less than PPP 2$ a day. Institutions and the vector of geographical variables were used as the independent variables. They find that poverty and institutions are negatively correlated. Countries with better institutions – control of corruption, regulatory quality, rule of law, government effectiveness, voice and accountability, and political stability – have smaller poverty rates, according to the authors. The geographical variable has no direct effects on poverty. The authors find that poverty increases in economies with weak regulatory systems or lack of law endowment by having low levels of income. However, poverty is affected via income distribution through corrup-tion, government effectiveness and political stability.

Other authors such as Perera and Lee (2013), studied the connection between GDPpc growth and poverty and found out that they have a negative relationship, meaning that pov-erty reduction may be caused by economic growth. In their results, there is evidence of a stronger institutional quality and better government stability that helps to decrease poverty levels. The authors prove that corruption levels, improves the democratic accountability and the quality of bureaucrats may increase poverty levels. The authors state that corruption in

(23)

16 Africa, Latin America and the Caribbean is different to the one in found in East Asia, since the last one is less unfavourable to growth and poverty. They find that government stability and law decrease poverty, and an increase in corruption index and democracy accountability index will increase poverty.

Azward (2018) also analyses the impact of corruption on poverty, using the HCR to measure poverty and the CPI to measure corruption, after controlling for other variables such as economic growth, inflation and unemployment. The author states that corruption affects indirectly poverty, because these levels of corruption lead to a smaller social spending and an increase in the levels of corruption could decrease the government spending on health and education. So, he argues that by decreasing economic growth, corruption in-creases and as a result poverty inin-creases.

There are also studies that investigate the effects of poverty on corruption. Poverty is affected by corruption through the reduced foreign and domestic investments, the distor-tion of a market and higher income disparities (Chetwynd et al., 2003). The institudistor-tions - either social, economic and political – are negatively affected by poverty that invites corrup-tion (Negin et al., 2010). The authors, Negin et al., state that factors such rural populacorrup-tion have a negative correlation with corruption – by increasing corruption because of increased rural population – and political freedom and stability have a positive correlation because if they are increased corruption decreases.

Justesen and Bjørnskov (2014) use OLS regressions to test the relationship between bribery and poverty. They use an afrobarometer to measure corruption, that contains ques-tions asking people if they pay bribes to have access to public services and to measure pov-erty, they ask people if is regular to have lack of access to household necessities that are basic to live. The authors find that poverty increases the regularity of individuals paying bribes to have access to services.

Unver and Koyuncu (2016) focus on the impact of poverty on corruption, after con-trolling for other variables such as trade openness, inflation and democracy indicators. In their research, they found some evidences that the higher rates of inflation and public debt levels are caused by the higher levels of corruption. The authors also find out that poverty strongly affects corruption: the higher the poverty levels the higher will be the corruption levels in a country.

Focusing on social welfare, Yildiz (2017) states that CPI has a negative relationship with inflation and Human Development Index and positive with unemployment, GDP per

(24)

17

capita and population. The author states that, since developing countries are unable to face

the social and economic problems, there is a problem regarding the distribution on prosperity to the population.

From the literature analysed, it is possible to conclude that corruption impacts pov-erty through different mechanisms, either through worsening income distribution or damag-ing institutions and public services. In addition, poverty can impact corruption, as increases the probability of paying bribes or having weak institutions.

The main focus of this dissertation is to study the impact of corruption on poverty and if education has some influence in this relationship. The next section outlines the litera-ture that studies the connection between corruption and education.

2.3.2. Corruption and Education

Chetwynd et al. (2003) claim that the number of children enrolled in schools and infant mortality are related with corruption, because if the public services are corrupt – due to the influence that corruption has on governance by reducing their capacity – and do not give importance to the expenditures on health and education, the levels of poverty increase, and the poor suffer the consequences. The authors state that in countries with high levels of corruption, the quality of the public services is worst. And because of that, the poor face the impacts through these channels. Due to the smaller governance capacity caused by cor-ruption, the public trust on government institutions is also affect. According to the authors, this makes the poor more vulnerable and their economic productivity suffer the impacts: reduced incentives to enter in productive activities. So, Chetwynd et al. argue that poverty is not produced by corruption; but corruption impacts poverty because it affects the economic growth and this in turn affects poverty.

Public officials with low levels of education tend to seek for bribes (Beets, 2005). As pointed by Beets, these public officials only understand the short-term personal profits they earn from bribes and do not care about the economic implications of corruption both in their countries and worldwide. Beets (2005) states that education is related to corruption, since countries with high values of perceived corruption are associated with relatively small enrolment in school, relatively low rates of literacy, and relatively classes with more students and only one teacher in primary school. So, low levels of investment in education and gov-ernment purposes may affect poverty in a country limiting the funds arranged by taxes.

(25)

18 as a ratio to GDP have a negative correlation. So, this means that the investments on educa-tion is lower when the levels of corrupeduca-tion are higher (Mauro, 1997; Tanzi, 1998). Other variables were also measured in Mauro (1997), such as transfer payments – welfare payments and social insurance – that are also negatively and significantly associated with corruption. But education is the only variable of public spending that has a correlation with corruption when the additional explanatory variable of the level of per capita income is used. The author states that education is a very important since is one of the determinants of economic growth. Also, Azward (2018) states that corruption indirectly affects poverty, because corruption lead to a smaller social spending and an increase in the levels of corruption could decrease the government spending on health and education.

De la Croix and Delavallade (2009) find that countries investments on housing and physical capital are higher when these countries have more predatory technology, and that the investments on education and health are smaller.

In Eicher et al. (2009) study, a model is presented that examines what problems block the growth of low-income countries with poor institutions, low human capital and high ine-quality. These authors defend that corruption can negatively affect education, but education also influences corruption. On one hand, corruption leads to the decrease of available in-comes and investment in education. On the other hand, education generates higher income, and increases the risk of corrupt politicians being detected and punished. Eicher et al. (2009) stated that in countries with high levels of education, governments tend to move away from corruption, since individuals have the incentive to be honest because they want to maintain their political power. In countries with low schooling, poverty is more likely to be trapped. Eicher et al. (2009) point out that one of the most corrupt countries in the world is Zimbabwe. As one of the least corrupt countries in Africa, they indicate that according to the Transparency International, Botswana is the country seen. They also note that, after the independence granted by the Botswana Democratic Party (BDP), this country had an invest-ment in education, health and public infrastructure, and then becoming one of the countries with high income per capita levels in Africa.

The authors also find some results concerning the development paths. The first one indicates that education expansion may be caused by institutional reforms that decrease cor-ruption levels. The second development path happens when human capital is increased by education subsides, decreasing future corruption on governments.

(26)

19 Eicher et al. (2009) also give the example of Latin American countries that are asso-ciated with poor institutions and corruption. These countries have intermediate education levels by the mid-20th_{century, and corruption rents were very high where political}

participa-tion was not strong enough to avoid these corrupt acparticipa-tions. Another example is East Asia in the 1950s, where the poverty trap was inevitable, since these countries were very populated, with weak resources and uneducated. But, increasing public education programs accompa-nied with high levels of per capita incomes were achieved in the decades after.

Another study that relates education and corruption is presented by Truex (2011), where the author considers some categories that are related to education as independent variables. The author considers some attitudes as behaviours of corruption in the public sector as the dependent variable. As a result, he finds that high level education is associated with less corrupt attitudes, meaning that there are positive effects of education on corruption by reducing the tolerance for corruption behaviours and this increases with high levels of schooling. In the type of corruption as favouritism, a stronger relationship is found between education and the smaller acceptance attitudes. Contrarily, the effect of education is smaller when bribery happens, since there are sympathetic respondents to this situation that are more educated.

Also, Kaffenberger (2012) points out that the effects of education on corruption are large, since education helps to decrease illegal behaviour, to increase social cohesion and civil responsibility. However, the author states that education can create more bribery opportuni-ties. As it is for common knowledge, developing countries have a tendency for corrupt edu-cation systems, where the students need to pay bribes for good grades or to move to the next grade level. Additionally, Kaffenberger argues that poor are likelier to pay bribes. As men-tioned by the author, if students pay bribes to have access to their education systems, they will believe that bribes are a solution and them practice bribes their own. In their results, the author finds that the higher the education level, the higher the probability of individuals participation in bribery in developing countries.

According to Harber (2002), education has a main role in poverty reduction. The author believes that the mechanism through which education can decrease poverty is democ-racy, but there is a need of a consciously form of education to foster democratic values and behaviours. There is evidence that authoritarian rules increase the levels of poverty through corruption, violence and wars. Harber (2002) points to the need of promoting democracy through education, although it would be a long walk. Another point of view is present by

(27)

20 Ogundele, Akingbade and Akinlabi (2012), stating that entrepreneurship education may lead to poverty alleviation because individuals achieve innovative skills that makes them job crea-tors rather than job seekers.

(28)

21

Chapter 3. Methodology

In this chapter we explain the methodology used in this research. First, we describe the model we use and then the variables and data.

3.1 The model

In order to fully examine the influence of corruption on poverty, this study uses a dataset comprising 81 low- and middle-income countries between 1998 and 2017 for which we have available information for at least three years concerning the dependent variable.

Our data set combines time-series and cross-sectional information, so we use panel data estimation since it incorporates the economic procedures and accounts for both heter-ogeneity across countries and dynamic effects (Gujarati, 2004). According to Greene (2012), there are different models in panel data such as the Pooled Regression, the Fixed Effects and Random Effects. Focusing on the last two, the author states that the fixed effects model assumes that the individual effect is unobserved and correlated with the dependent (or ex-planatory) variables. For the random effects, the unobserved individual heterogeneity is un-correlated with the dependent (or explanatory) variables. Gujarati (2004) point out that the fixed effects exist because the individual intercept does not change over time although the intercept may be different between individuals, being time invariant. According Gujarati (2004), when the number of cross-sectional units is higher than the number of periods (as is our case), FEM may be adopted.

The econometric model used can be described as: 𝑌𝑖𝑡= 𝛽1𝑋𝑖𝑡+𝛽2𝑍𝑖𝑡+𝛼𝑖+ t +𝑢𝑖𝑡,

where i represents the country (i = 1, …, 83) and t represents time (t = 1998, …, 2017). 𝑌𝑖𝑡 is the dependent variable and refers to a measure of poverty (Poverty Headcount Ratio) of country i at time t; 𝛽1 is a vector of coefficients associated with the explanatory variables; 𝑋𝑖𝑡

is the vector of explanatory variables (Corruption, measured by CPI, and Education, meas-ured by School enrolment primary (% gross), School enrolment secondary (% gross) and Government expenditure on education (% of GDP)), defined for each country i at time t; 𝛽2 is a vector of coefficients associated with the control variables; 𝑍𝑖𝑡 is the vector of control

variables, defined for each country i at time t; 𝛼𝑖 and t are the unobserved country and time

(29)

22 (2004), we have an unbalanced panel as the number of observations among the panel mem-bers is different. To estimate the model, we use eViews software package, version 10.

3.2 Data

This study aims to analyse the impact of corruption on poverty focusing on develop-ing countries, so we consider a dataset of countries with low and middle income.11

Considering the World Bank12_{classification, the structure of the sample is the}

follow-ing (Table 3):

Table 3 – Structure of the sample by income level

Threshold GNI per capita (current US$) Number of countries

in the sample

Low-income < 995 17

Lower-middle income and upper-middle

in-come

996 - 3,895

64 3,896 - 12,055

High-income > 12,055 0

Source: World Bank13

On Figure 1, we can see the evolution of the average of Poverty HCR between 1998 to 2017 from the different classifications of countries made by the World Bank – low-in-come, middle-income and both.

11_{We adopt the World Bank’s classification that uses the GNI per capita calculated with the Atlas method (in} current US dollars) for 2017. The World Bank aggregates economies in four categories: high income, upper middle income, low middle income and low income. Available at https://data-helpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups (accessed on 26/03/2019)

12 https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups (accessed on 26/03/2019)

13 _{https://blogs.worldbank.org/opendata/new-country-classifications-income-level-2018-2019}_{(accessed on:} 26/03/2019)

(30)

23

Figure 1 – Poverty HCR in low- and middle-income countries, 1998-2017

The low-income countries in Figure 1 present higher values of the Poverty HCR, which makes sense because these countries are the poorer countries in the world. When considering the total sample, the Poverty HCR decreases during the period, while it increases in the low-income sample. From the figure, we see a discrepancy between the low-income countries and the total sample, this happen because there are only 17 low-income countries and 66 middle-income countries that have more impact on the average of the full sample.

Figure 2 shows the evolution of CPI in our sample, for the same period.

0,0 10,0 20,0 30,0 40,0 50,0 60,0 70,0 80,0 90,0

(31)

24

Figure 2 – CPI in low- and middle-income countries, 1998-2017

As we can see in Figure 2, the control of corruption is higher- in middle-income countries when compared with low-income countries, particularly in recent years. The mid-dle-income countries reached in 2017 a CPI of 31, on average, that compared with 89 for New Zealand14_{, the highest one. For low-income countries, CPI is lower, which makes sense}

because these countries have smaller control of corruption levels which means that they are more corrupted than the middle-income ones.

Next, we describe the variables. First, we will present the definition and justification of the dependent variable. Then, we will do the same for the independent variables (explan-atory and control variables).

Dependent variable

Our dependent variable is poverty. Considering the literature (e.g. Alvi and Senbeta (2011), Azward (2018), Chong et al. (2009), Perera and Lee (2013)), we use the Poverty Head-count Ratio at $1.90 a day (2011 PPP) as a measure of poverty. This indicator measures the proportion of population below the monetary poverty line (Mabughi and Selim, 2006). Since

14 _{https://www.transparency.org/news/feature/corruption_perceptions_index_2017} _(accessed _on 26/06/2019) 20 22 24 26 28 30 32 34 36

(32)

25 this measure is the simplest measure and most common one in the literature analysed, we choose it instead the Poverty Gap or the Squared Poverty Gap, in order to fully compare our results with other studies. The data is taken from the World Development Indicators of the World Bank (World Bank, 2019).

Explanatory and control variables

The explanatory variables are corruption and education. There are different control variables considered for controlling the influence of other variables on poverty:

- Corruption, measured by the Corruption Perception Index (CPI) from the Trans-parency International from 1995 to 2017. As mentioned before, corruption is very hard to measure, so we choose this index since it is the most widely used in the literature examined (e.g. Azward (2018), Rahayu and Widodo (2012)) and it measures corruption nationally. This measure varies from 0 to 100, being 0 the highly corrupt countries and 100 the highly clean countries. As pointed by Rahayu and Widodo (2012), corruption directly affects the poor by increasing the prices of public services, leading to a smaller quantity and quality of these public services and making difficult to the poor to have access to services such as health, education, sanitation, … Also, corruption affects poverty indirectly by lowing the economic growth, increasing income inequality and by distorting the public expenditure allocation (Azward, 2018).

- Education, measured by School enrolment primary (% gross) and School enrolment secondary (% gross), each of them corresponding to the level of education shown. This measure is taken from the World Bank (2019), but the source is from UNESCO Institute for Statistic. According to the World Bank, primary education gives children skills such as read-ing, writing and mathematics skills needed, and secondary education completes the provision of basic education and offers skill-oriented instruction with teachers more specialized. Ogundele et al. (2012) point out that entrepreneurship education is effective on the poverty reduction. Since the expenditures on education are also a measure of education and in order to see the investment that government have on this variable, the Government expenditures on education (% of GDP) are included. This variable includes expenditure funded in current, capital and transfers, and it is available at World Bank World Indicators database (World Bank, 2019). According to the literature (e.g. Azward, 2018, Gupta et al., 2002), we shall expect that the lower the social spending is, the higher is poverty.

(33)

26 The control variables are chosen taking into consideration the literature analysed in previous chapter:

- GDPpc, at 2011 international Purchasing Power Parity (PPP) dollars is retrieved from the World Banks World Development Indicators database (World Bank, 2019). Con-sidering the studies of Azward (2018) and Perera and Lee (2013), we use this measure to control for the level of economic development in a country. Perera and Lee (2013), have showed in their results that the relationship between GDPpc and poverty is negative, meaning that poverty reduction is related with economic growth.

- Current health expenditure (% GDP), includes the goods and services of healthcare that are consumed each year, and it is available at World Bank - World Indicators database (World Bank, 2019). According to Azward (2018) and Gupta et al. (2002), decreasing the social spending on health increases poverty.

- Trade openness, refers to the sum of exports and imports of goods and services as a share of GDP and it is retrieved from the World Bank -World Indicators database (World Bank, 2019). Perera and Lee (2013) point out in their conclusions that one of policies that help to reduce poverty is international trade openness. Alvi and Senbeta (2011) show that trade and poverty have a negative correlation, which means that countries with higher open-ness experience poverty reduction.

- Gross Capital Formation (GCF) consists in investments in fixed assets as a per-centage of GDP and is gathered from the World Banks World Indicators database (World Bank, 2019).

- Urban Population, referring to individuals that live in urban areas (as a percentage of total population), is sourced from the World Bank World Indicators database (World Bank, 2019). According to Negin et al., (2010), the likelihood of being poor in developing countries is higher in rural areas than in urban areas and the severity of poverty is also higher in rural areas.

- Gini index, which measures the distribution of income among individuals or fami-lies in an economy, deferring from 0 (perfect equality) to 100 (perfect inequality). Gupta et al. (2002) argue that Gini and income growth of the poor are negatively correlated. As pointed out by Negin et al., (2010), higher income inequality reduces economic growth and worsens poverty.

(34)

27

Table 4 – Summary Statistics

Variable Mean Median Max Min _DeviationStandard Source

HCR count Ratio (% Poverty

Head-population) 17.775

6.100 _86.000 _0.000 _18.280 World

Bank

CPI Perception In-Corruption

dex 31.859 31.000 67.000 4.000 9.016889

Trans-parency

Interna-tional

EDUCPRI ment, primary School

enrol-(% gross)

101.32

5 102.547 165.645 29.023 17.938 World Bank

EDUCSEC ment, second-School

enrol-ary (% gross) 65.160 71.141 126.054 5.210 26.824 World Bank EXPEDUC Government expenditure on education, total (% of GDP) 4.220 4.031 9.662 0.000 1.563 World _Bank GDPPC GDP per cap-ita, PPP (con-stant 2011 in-ternational $) 6790.7 69 5676.682 26824.08 540.095 5293.181 World Bank

EXPHEA expenditure (% Current health

of GDP) 5.544 5.279 13.677 0.966 1.858

World Bank

GCF formation (% Gross capital

of GDP) 24.448 23.193 73.777 0.000 8.795

World Bank

TRADE Trade (% of _GDP) 74.950 68.519 220.407 16.439 32.860 World _Bank

URBPOP Urban popula-tion (% of

to-tal) 47.775 47.334 90.747 7.830 19.361

World Bank

(35)

28

Chapter 4. Do corruption and education impact poverty in developing

countries?

In this chapter the results and the estimation of the models are present, as well as their analysis. The sample corresponds to 81 countries that are considered by the World Bank as low-income, low middle-income and upper middle-income and for which we have infor-mation for the dependent variable – Poverty Headcount Ratio – at least for 3 years. We will first estimate the total sample and then, we consider a subsample with the low-income coun-tries. The following results are own work and elaboration.

We start by computing the correlation coefficients between all pair of variables (Table 5). The correlation coefficient between the Poverty Headcount Ratio and CPI, is negative and statistically significant, which means that the higher the control of corruption (CPI), the lower the poverty (HCR).

From the analysis of the table, we can state that the correlation between the variables is low, with all correlation coefficients lower than 0.6. However, there are some exceptions such as the correlation between HCR and School enrolment secondary (% gross) and the correlation between Urban Population and School enrolment secondary (% gross), with a value of -0.729 and 0.686, respectively.

(36)

29

Table 5 – Correlation Matrix

Note: p-value in parenthesis; the significance at 1% (***), 5% (**) and 10% (*)

HCR CPI EDUPRI EDUSEC EXPEDUC GDPPC EXPHEA GCF TRADE URBPOP GINI

HCR 1.000 --- CPI -0.140** (0.030) 1.000 --- EDUPRI 0.022 (0.734) -0.274*** (0.000) 1.000 --- EDUSEC -0.729*** (0.000) 0.264*** (0.000) 0.011 (0.871) 1.000 --- EXPEDUC -0.150* (0.021) 0.202*** (0.002) 0.064 (0.321) 0.232*** (0.000) 1.000 --- GDPPC -0.566*** (0.000) 0.400*** (0.000) -0.137`` (0.034) 0.583*** (0.000) 0.040 (0.543) 1.000 --- EXPHEA -0.073 (0.264) 0.125* (0.054) -0.280*** (0.000) 0.250*** (0.000) 0.510*** (0.000) -0.172*** (0.008) 1.000 --- GCF -0.173*** (0.008) -0.202*** (0.001) 0.033 (0.612) 0.113* (0.082) 0.008 (0.898) 0.079 (0.224) -0.094 (0.150) 1.000 --- TRADE -0.201*** (0.002) -0.042 (0.517) -0.179*** (0.006) 0.059 (0.363) 0.277*** (0.000) 0.055 (0.395) -0.009 (0.894) 0.246*** (0.000) 1.000 --- URBPOP -0.585*** (0.000) 0.313*** (0.000) 0.052 (0.426) 0.686*** (0.000) 0.086 (0.188) 0.625*** (0.000) 0.097 (0.134) -0.130** (0.045) -0.224*** (0.001) 1.000 --- GINI 0.088 (0.173) 0.497*** (0.000) -0.033 (0.609) -0.015 (0.817) 0.010 (0.881) 0.241*** (0.000) 0.004 (0.952) -0.426*** (0.000) -0.236*** (0.000) 0.307*** (0.000) 1.000 ---

How Corruption Impacts Poverty in Developing Countries? The Role of Education